Can automated vehicles recognise when they are uncertain? New partner project will find out!
Unusual objects, unexpected behaviour and bad weather can challenge automated driving systems. A partner project newly linked to SAFER's knowledge hub explores how AI can recognise its own uncertainty and help vehicles respond more safely.
An automated vehicle needs to assess how confidently it understands a traffic situation. Recognising uncertainty could help it adapt its driving, for example by slowing down when the situation is difficult to interpret.
The project Risk assessment in vision-language-action models for autonomous driving brings together Linköping University and Zenseact. It studies AI models that combine visual information, language and driving actions, testing ways to measure their uncertainty and assess collision risk.
The team will also interview experts from industry and academia to identify situations where this information could support safer automated driving.
Funded by Vinnova through the FFI program, the project runs until April 2027.
We welcome the project to SAFER’s partner project portfolio and look forward to sharing knowledge across the network. We hope its findings will contribute to safer and more reliable automated vehicles
